Merged Discussion of femr's video data analysis

:rolleyes: I "guess" you are not going to choose to justify why you consider mine only "probably" more accurate ?

You could have said "definitely" more accurate, whilst still complaining about error threshhold, yet you have chosen not to.
I've too often run into arguments that looked convincing and were dismounted later. I am careful with such things now.


It doesn't gain respect, nor credibility.
Recklessness doesn't gain respect, nor credibility. I am more careful and therefore I also consider the possibility that your analysis is wrong.


The fact is, that your assertion of "chaotic and unpredictable" has two primary implications. One you'll "like", the other you won't.
Why are you introducing the topic of implications in this thread? And I'm the one accused of gish-gallop? :rolleyes:

I'll commit the "gish-gallop" of answering, though.


1) If even 0.00001 degree of error renders the subsequent simulation "chaotic and unpredictable", then the NIST simulations are completely worthless chaotic and random number generators, especially if you consider that the scale of error of one impact parameter (of several that are wrong) is around nine hundred thousand times larger than your offered "threshold".
The NIST simulations are not worthless, but they are chaotic. They offer a starting point. NIST had no data of core damage to work with; they couldn't do any analysis without it. They had two options: find a plausible starting point to see where it led, or not do the analysis. Along the way, they matched the data to the observables where possible. The final result matches their observations, so theirs was as good a starting point as many others.

But the errors in column location, non-uniformity of column density, weld resistance, and so many other factors, cause that performing a simulation to determine the damage to give results basically guaranteed to differ from real life results. Even two real life impacts with the very same conditions to within my example error above would produce different results.

By the way, the threshold you're confident with in your analysis is also a hundred thousand times bigger than my example error!!!1!11!eleventy!


Ah, switchback to a separate "argument". Canter :)
:eye-poppi What separate "argument"?
I do hope you're not going to try and foist a manual by hand lens correction upon anyone.
I do hope you're not going to try and foist a manual by hand lens correction upon anyone.
Wait until I provide it and judge then. But out of curiosity, what would be wrong with that?
But out of curiosity, what would be wrong with that?
I'll wait until you provide it and judge then ;)
I'll wait until you provide it and judge then ;)
So you won't tell me what's wrong with that approach?
So you won't tell me what's wrong with that approach?
Why would you ask ?
Why don't you answer? That paper hints that it could be done:

When determining the lens parameters, all programs rely on the same paradigm: the ideal perspective mapping should map real world straight lines to straight lines in the image. So if a set of points P0, P1, ..., Pn is known to lie on a straight line, their images p0, p1, ..., pn should also fall onto a straight line [...]​

But rectification of lines is not the method I've used, anyway.
Why don't you answer?
I've provided you with a full paper about correcting lens distortion. What extra information do you need now ?!
I've provided you with a full paper about correcting lens distortion. What extra information do you need now ?!
My question was: "out of curiosity, what would be wrong with that [= a manual by hand lens correction]?"
My question was: "out of curiosity, what would be wrong with that [= a manual by hand lens correction]?"
Ah, switchback to a separate "argument". Canter :)
:dig:

Compare the highlighted sentences.


And did you learn nothing from the document supplied and my numerous other responses ?
Yes, I learned that there are better open source software packages for lens correction than the one I have been using, and I intend to repeat my adjustment with them.


Short answer...inaccurate. Lens parameters are given to 6dp for a reason. If you can't achieve accurate correction you might as well not bother.
Wrong. If I can determine the error, I can account for it in the final result. Introducing a known amount of error is much better than not knowing what error you have at all.


I'm sure you're aware that lens distortion varies between barrel, complex and pincushion types as zoom level is changed ?
Yes, and I'm sure you read this sentence?
- The final FOV is similar to the FOV in the collapse view, therefore the lens distortion is similar too.


How were you planning on separating change in perspective distortion during zooming at the same time ?
Seems you haven't understood. I'm only using two frames, pre-zoom and post-zoom. I assume negligible lens distortion in the center of the image for the not zoomed view, and adjust the distortion of the zoomed view to match the magnified view. That will take me much closer to the actual lens distortion than without any correction at all. I can determine a plausible error from there, which will be substantially reduced with respect to not applying lens correction at all.


How are you planning on performing lens correction upon the Dan Rather footage, as you're planning on using both... ?
I plan to perform no lens correction, and instead determine an error band resulting of that omission. ETA: Unless I can find another zooming video similar to the one for Cam #3. But I suspect it won't be so close to the center if that's the case, which will ruin it.
 
Last edited:
d) pgimeno has generated his own velocity and acceleration derivative plots
e) pgimeno's acceleration plot includes an over-g period
Irrelevant, given that I made my acceleration plot deliberately to match yours as closely as possible, in order to show what actual velocity and acceleration curve your graph corresponds to. I would have used your exact data or methods had you provided them; instead it was a trial-and-error attempt to match your graph.
 
Last edited:
We're talking precision here, not correctness.
The NIST Flight 175 impact orientation and trajectory parameters are wrong. Ther are not correct. They are incorrect.

Let me put it in radians
Why the hell would you choose to do that ? What motive does it serve ? Try and make the numbers look smaller ?

femr2 is saying that the error in NIST's estimation is about 0.166 radians
No, he isn't.

I've provided values for several separate orientation parameters in degrees.

If you'd bothered to digest details provided you wouldn't make such basic errors.

and he gives an estimation with an error of... wait, how much? Oh, yes, he doesn't give it
Nonsense. I've provided 2 bounded error ranges, for camera placement accuracies within 100ft and within 10ft.

How about...
Video rotoscoping is used to verify the impact orientation. This calculation allows us to determine the accuracy of the viewing angle. The viewing angle accuracy in turn directly defines the accuracy of the rotation of the tower model, and the 3D aircraft model.

It has been previously shown that in order for the aircraft model to match video footage, it must be rotated 9.5 degrees from the angle used by NIST.

If our draft camera location is within 100ft of the actual camera location, the rendered view, specifically the rotation of the tower and aircraft, is therefore within around 1.7 degrees of the actual value.

How far off would the camera location have to be to generate a 9.5 degree error ?

Perpendicular distance = 3342 * tan(9.5 degrees) = 559.1166ft.

In reality we can place the camera with an accuracy well within 100ft, as not only have the occupants of the ferry upon which the footage was taken stated their specific positions, but there is also additional photographic footage available from the same time aboard the same ferry, from a very slightly different viewing position, taken by a different person. This allows additional clarification to be performed.

Beyond multiple footages, from essentially the same location, is the fact that other buildings within the viewing frame allow for parallax-based positioning to be performed, significantly increasing camera viewpoint accuracy.

I'm confident that camera positional error is up to around the 10ft mark.

but claims (without proof)
Jeez. See above.

that it's less than 0.018
Say wha' ?

1. Femr2 claims the NW corner went over g.
LOL. If you're going to continue trying to justify your nonsense...quote me "claiming the NW corner went over g".

2. As a proof of that claim, he shows a graph where the NW corner went over g.
What claim ?

And any suggestion of over-g behaviour from me is not based upon a single graph son, but poly(10, poly(50), symmetric difference (13), "old" SG, 1-pass SG, linear regressions, NIST data, Chandler data, COMMON SENSE, ... :rolleyes:

You are REALLY boring me.

You "believe" that the (ridiculous wandering) point NIST tracked reached approximately gravitational acceleration for around 2.25s, of course...

Now, it is blatantly obvious to anyone with half a noodle that the NW corner descent lags behind that point, and is "pulled" by other elements of the building, later descending at about the same rate as the NIST "point". Therefore the NW corner MUST exceed g for a period of time.

Frankly I don't care if you don't want to accept the presence of the indicated over-g period, as shown by femr2, NIST, Chandler, ...

You are one person on the internet who has painted your intent very clearly, namely...

"No, I won't put a threshold on the error to consider something correct or not."

Yawn.
 
Why are you introducing the topic of implications in this thread?
Because you tried to suggest that even a 0.00001 degree error would have a chaotic and unpredictable effect on simulation behaviour :rolleyes:

The NIST simulations are not worthless, but they are chaotic. They offer a starting point. NIST had no data of core damage to work with; they couldn't do any analysis without it. They had two options: find a plausible starting point to see where it led, or not do the analysis.
What would have changed if they had not used incorrect impact parameters ?

Along the way, they matched the data to the observables where possible.
Where do they apply FEA element state based upon observables to the model ? References please.

The final result matches their observations
:jaw-dropp THAT statement will make MT's brain explode.

Even two real life impacts with the very same conditions to within my example error above would produce different results.
Of course, with current primitive software, and even more so with the hacked ANSYS installation NIST used. How different ? How different would it have been with an error NINE HUNDRED THOUSAND times larger ? ;)

By the way, the threshold you're confident with in your analysis is also a hundred thousand times bigger than my example error!!!1!11!eleventy!
You quite sure about that ? :rolleyes:

What separate "argument"?
From WTC2 aircraft impact orientation to WTC7 trace data.

Compare the highlighted sentences.
Yes, you appear to have learned nothing from the detail in the document provided, or you would not keep asking such a basic question.

Yes, I learned that there are better open source software packages for lens correction than the one I have been using, and I intend to repeat my adjustment with them.
Okaaaay...

Wrong. If I can determine the error, I can account for it in the final result. Introducing a known amount of error is much better than not knowing what error you have at all.
It's going to be funny as hell when you're "finished". Can't wait. You suffer from extreme hubris you know. I stand by my assertion ;)

Are you sure you understand why I wrote it ?

Seems you haven't understood.
Disagreeing with your assertions about your newly found "toy", lens distortion correction, does not mean I do not understand.

That will take me much closer to the actual lens distortion
:eek: There's some logic to your "idea", but the problem really is that you have no information about the lens. You don't know at what focal lengths it exhibits any combination of barrel, complex or pincushion. You don't know the transition behaviour between barrel, complex or pincushion during zoom. But you just go ahead, don't listen to me. I know absolutely nothing about lenses, optics, video, imaging, .... :rolleyes:

I can determine a plausible error from there, which will be substantially reduced with respect to not applying lens correction at all.
How are you intending on "correcting" lens distortion ?

I plan to perform no lens correction [for Dan Rather]
Interesting.
 
The NIST Flight 175 impact orientation and trajectory parameters are wrong. Ther are not correct. They are incorrect.
Yours are wrong too.


Why the hell would you choose to do that ? What motive does it serve ? Try and make the numbers look smaller ?
Emphasize that the issue at hand is precision, not correct vs. incorrect.


that it's less than 0.018
Say wha' ?
femr2 in the911forum.freeforums.org said:
How many camera locations do we need to conclusively determine the impact angle(s)?
I think we're well within one degree at the point of impact
http://the911forum.freeforums.org/carmen-taylor-photo-analysis-t196.html#p4198

1 degree is the best value I've seen on the error of your impact angle estimation. Do you have any better one? And where is it proved?


LOL. If you're going to continue trying to justify your nonsense...quote me "claiming the NW corner went over g".
I won't bother looking among your 3500+ posts. Suffice to show your admission a few pages ago:
2) You have claimed over-g. Several times during the last years.
I have indeed.


Because you tried to suggest that even a 0.00001 degree error would have a chaotic and unpredictable effect on simulation behaviour :rolleyes:
Indeed, and the times I've had to respond to your attacks, which you've used to accuse me of "gish-galloping" also had their origin in the ongoing discussion. In this case, that claim arose when you asked me to quantify what error is acceptable.


What would have changed if they had not used incorrect impact parameters ?


Where do they apply FEA element state based upon observables to the model ? References please.


:jaw-dropp THAT statement will make MT's brain explode.


Of course, with current primitive software, and even more so with the hacked ANSYS installation NIST used. How different ? How different would it have been with an error NINE HUNDRED THOUSAND times larger ? ;)
I won't discuss that here.


You quite sure about that ? :rolleyes:
By the above, yes. Again, if you have a better estimate of the error in your impact angle, tell me.


From WTC2 aircraft impact orientation to WTC7 trace data.
They are two running topics. You invited in another thread to discuss WTC2 aircraft impact orientation here, and I did have things to comment on that while the WTC7 trace data was being discussed here. I'm not introducing new topics on purpose to divert attention from previous topics, which is the definition of a "gish gallop". You typically respond separately even to fragments of the sentences you quote, which tends to multiply the topics. I try to keep my discussion focused on a single topic, by quoting the minimum possible and responding only to it. With you, and because of your ramification of topics, that soon proved to be impossible.

But to sum up, if you didn't want to keep both topics running at the same time, maybe you shouldn't have invited people to discuss it here until we were done. Your implied accusation of gish-galloping was dishonest.


Yes, you appear to have learned nothing from the detail in the document provided, or you would not keep asking such a basic question.
What I am doing is indeed adjusting the lens correction by hand, but using a known "good" image as the pattern to match it against, instead of doing it blindly. The match has also sub-pixel accuracy.

The reason I asked is because it's the method I'm using and I foresaw you could be using that as another attack. The problem of adjusting by hand is not having a good reference image to match against, but I do.


Are you sure you understand why I wrote it ?
No, I don't understand why you wrote that because I don't think it applies to my method. As I said, I'm using an area at the center of the zoomed out image, which precisely because of being at the center, has presumably very little lens distortion. I am considering the lens distortion at the center negligible, therefore I don't care at all what kind of lens distortion applies at that zoom level because I'm going to ignore it. The lens distortion in the zoomed in image is the one I care about and will correct, and then, since it has a very similar FOV as the collapse video, use the values found for that image to apply lens correction to the collapse video frames.
 
Yours are wrong too.
Of course. A hell of a lot less wrong than NIST though.

I note your acceptance that the NIST Flight 175 impact orientation and trajectory parameters are wrong. Good.

Doesn't sit well with your prior statement : I won't put a threshold on the error to consider something correct or not.

Emphasize that the issue at hand is precision, not correct vs. incorrect.
I'm clearly asking you why you chose to recalculate values provided in degrees to values in radians.

1 degree is the best value I've seen on the error of your impact angle estimation. Do you have any better one?
I suggested camera accuracy within 10ft... 0.17 degrees. You can add error in aircraft model rotoscope orientation to that. About 1 degree sounds fine. Could be a bit less.

And where is it proved?
LOL.

Suffice to show your admission a few pages ago
:) Context and inference. I have indeed claimed that multiple methods of determining acceleration trend profile indicate over-g behaviour.

There's a subtle difference.

Indeed, and the times I've had to respond to your attacks, which you've used to accuse me of "gish-galloping"
Say wha' ? Where have I accused you of "gish-galloping" ?

In this case, that claim arose when you asked me to quantify what error is acceptable.
And you respond with 0.00001 degrees ? :jaw-dropp

I'm good, but ... :rolleyes:

I won't discuss that here.
Fine.

By the above, yes. Again, if you have a better estimate of the error in your impact angle, tell me.
My response was due to you bizarrely stating a value in radians (without bothering to denote it as such).

They are two running topics.
I know. You switched from one to t'other.

I suggest you stick to one "analysis" within a post.

I'm not introducing new topics on purpose to divert attention from previous topics
I didn't say you did.

which is the definition of a "gish gallop"
See above.

But to sum up, if you didn't want to keep both topics running at the same time, maybe you shouldn't have invited people to discuss it here until we were done.
ROFL. See above.

Your implied accusation of gish-galloping was dishonest.
LOL. I simply said "canter" :) A jump from one topic to another within the same post. Messy.

What I am doing is indeed adjusting the lens correction by hand, but using a known "good" image as the pattern to match it against, instead of doing it blindly.
How do you know your "good" image is ? I assume you are aware that complex type distortion can be maximised near the focal point ?

The match has also sub-pixel accuracy.
:eye-poppi Interesting.

The reason I asked is because it's the method I'm using and I foresaw you could be using that as another attack.
:rolleyes: There really is very little of my dialogue with you that could validly be classed as "attack". I laugh quite a lot at things you say, but that's your own fault I'm afraid.

The problem of adjusting by hand is not having a good reference image to match against, but I do.
See above.

No, I don't understand why you wrote that because I don't think it applies to my method. As I said
I read what you said.

I'm using an area at the center of the zoomed out image, which precisely because of being at the center, has presumably very little lens distortion.
That. Might be true. Might not.

I am considering the lens distortion at the center negligible, therefore I don't care at all what kind of lens distortion applies at that zoom level because I'm going to ignore it.
Hmm.

The lens distortion in the zoomed in image is the one I care about and will correct, and then, since it has a very similar FOV as the collapse video, use the values found for that image to apply lens correction to the collapse video frames.
:jaw-dropp

A reminder...

I strongly suggest that you think carefully about your response.


Look at your last sentence there, and tell me what is catastrophically wrong with it.
 
Just in case the usefulness you refer to applies to the claim of over g, well, if we don't know whether it's true or false, how useful can it ever be for any purpose? What is a false statement useful for, other than as an example for a lesson in logic maybe? The possibility that it's false destroys any usefulness in it.
Pgimeno, this last point, regardless of the rest of your post, is in my opinion incorrect. Even if femr2's analysis has an error margin that puts the acceleration above g to a certain confidence factor for the error, the nature of the probability distibution still generally allows for some tiny possibility that he is wrong. Therefore even with a possibility it is false, that does not destroy the usefulness of it. And regardless of that, the possibility that he is right means that a theory that encompasses both his acceleration curve AND a sub-g acceleration curve trumps one that can only explain g and under accelerations.
 
So you decline to apply a binary status to an analogue value? Says the person who took the reverse position when calling ozeco wrong. Mmm....
scratch.gif
What position was that, exactly? Can you quote?
Of course I could but I will not play your derail and evasion games. Whether you intend the "Gish Gallop" or not pressing people like me who are trying to help you into wasting time and effort responding to irrelevancies, untruths, distortions or already well answered questions is the essence of the Gish Gallop.

So whether you intend it or not you would be better served by desisting from the practice.
 
Pgimeno, this last point, regardless of the rest of your post, is in my opinion incorrect. Even if femr2's analysis has an error margin that puts the acceleration above g to a certain confidence factor for the error, the nature of the probability distibution still generally allows for some tiny possibility that he is wrong. Therefore even with a possibility it is false, that does not destroy the usefulness of it. And regardless of that, the possibility that he is right means that a theory that encompasses both his acceleration curve AND a sub-g acceleration curve trumps one that can only explain g and under accelerations.
Thanks. That can be a basis for discussion and that might cause me to end up backing off on certain claims. That's not the way I've been presented the arguments so far, though.

I think we already have a reasonable hypothesis (theory, if you prefer) that explains over-g. What I wasn't sure is if over-g was correct. I'm still not fully convinced, but much closer than before.
 
With all the words you have expended in this thread, I think it would be more useful to make fewer points, better supported.
Nah. Abandon the aim of help pgimeno - the void between our reasoning processes is too wide to bridge. :o

Thanks for the comment though - I was about to abandon the pgimeno topic. I'm simply not into point by point building block arguments when we have an expert like femr on his own topic to tackle it that way. My input would be redundant at that level so I don't try. I find it easier to put forward the framework then decide which blocks belong where. It's a hell of a lot faster but....horses for courses I suppose.

I'm sorry that you see the need for "better supported" - but lets be sure which topic you are referring to. I've spent a lot of effort trying to bridge to the pgimeno topic. Let's forget that one. Go back a week or two where what I regard as the main issue was under discussion.

My main area of technical interest in this thread - stated many times with AFAICS zero support (:o) is:
"Do femr's video analyses produce data which is accurate enough to be useful?"
...I suggest that is the implicit main topic. Do you agree, disagree or suggest alternate wording?

Few enough words? Does it need support? ;)

(COMMENT most of what I have posted on that topic is near pure reasoning - or left open ended for modifying comments - zilch - zero - nil - there really is bugger all interest evident.

....however what support should it need?)

Now let's try step by step (BTW this argument has been put several times and essentially ignored leaving me out on my lonesome except for femr.)

The sort of measurements that femr is making will be subject to errors in three main areas viz (i) Noise; (ii)geometry (perspective, tilt, etc); AND (iii) Artefacts of the mathematical processing. Do you agree? Disagree? Suggest that I've missed something? Want to say it differently?

There that is the objective and step one of the argument. Do we have the basic foundation?

Do you want less words? More support? :)

The next step is a biggie. :boxedin:
 
Last edited:
I note your acceptance that the NIST Flight 175 impact orientation and trajectory parameters are wrong. Good.
Yes, everything measured is wrong. Which doesn't give much information.


Doesn't sit well with your prior statement : I won't put a threshold on the error to consider something correct or not.
It does. The only logical threshold is error=0. That's unattainable therefore everything is incorrect.


Emphasize that the issue at hand is precision, not correct vs. incorrect.
I'm clearly asking you why you chose to recalculate values provided in degrees to values in radians.
I did so to: Emphasize that the issue at hand is precision, not correct vs. incorrect.

If you ask once again the response won't vary the next time.


Say wha' ? Where have I accused you of "gish-galloping" ?
I thought your responses to Ozeco and the canter insinuation implied that. My apologies. ETA: English is not my first language, and when I looked up canter in a dictionary it translated to "medio galope" (half gallop). I took that as an insinuation of a Gish-gallop accusation.


And you respond with 0.00001 degrees ? :jaw-dropp

I'm good, but ... :rolleyes:
:rolleyes: Look at my response in context.


My response was due to you bizarrely stating a value in radians (without bothering to denote it as such).
Those of us who can interpret things in their context will understand that the values in this whole paragraph are in radians:

Let me put it in radians. femr2 is saying that the error in NIST's estimation is about 0.166 radians, and he gives an estimation with an error of... wait, how much? Oh, yes, he doesn't give it, but claims (without proof) that it's less than 0.018. To me, that counts as precision.
The clear (to a normal person) hint is that I'm comparing, and I say at the beginning that I'm putting it in radians. If you split the paragraph into completely segregated sentences, you miss the context. But that's your fault, not mine.


I suggest you stick to one "analysis" within a post.
I tend to minimize posts, but if that bothers you, I'll try from now on.
 
Last edited:
With all the words you have expended in this thread, I think it would be more useful to make fewer points, better supported.

I'd tell you what I mean but then you'd know so I'll just make elliptical remarks vaguely indicating that you aren't smart enough to figure out what I'm actually talking about. This allows me to maintain my smug attitude.:boxedin::mad:
 
How do you know your "good" image is ? I assume you are aware that complex type distortion can be maximised near the focal point ?
I just assume it is. Maybe it isn't, but that's just an attempt to reduce the error band. In the end, quantification of the error is what counts. Even if it's wrong, it will only mean that I won't reduce the error that much, but that's OK as long as I can track the error.


:rolleyes: There really is very little of my dialogue with you that could validly be classed as "attack". I laugh quite a lot at things you say, but that's your own fault I'm afraid.
Most of your laughs have been proven unjustified. Which makes me laugh in turn.


:jaw-dropp

A reminder...

I strongly suggest that you think carefully about your response.


Look at your last sentence there, and tell me what is catastrophically wrong with it.
I don't know. I see nothing clearly wrong there. Do you mean that a small difference in FOV can have enormous effects in lens distortion?

Here are the two views in question, the zoomed in one and the collapse video one:
http://www.formauri.es/personal/pgimeno/xfiles/11-s/LensViewComparison.gif

There's some difference in FOV, but I assume it's small enough to not matter much for lens correction.

Here's the zoomed out frame:
http://www.formauri.es/personal/pgimeno/xfiles/11-s/Cam3Far.png

That's the one I use to magnify a region at the center. Here's the comparison of the scaled area with the uncorrected zoomed view:

http://www.formauri.es/personal/pgimeno/xfiles/11-s/LensWithout.gif

and here's the comparison with the corrected zoomed view, still using Gimp 2.6 built-in lens correction tool, which only has two parameters and center adjustment:

http://www.formauri.es/personal/pgimeno/xfiles/11-s/LensWith.gif

The match is actually quite worse than what I actually obtained. I'm having a problem with Gimp's scale tool, where the size and the position vary between the preview (which is what I'm using for matching) and the final result, and it's difficult to compensate for. Maybe I can take two screenshots showing the previews, but I'll probably first take a look at Gimp's code to see if it's fixable. That'll be tomorrow, most likely.

In the better image, the windows below WTC7 still show deviation. The ones at the corners match pretty well, and the buildings at the right do too without the apparent movement visible in that last image. But I'm not going to try to fix that with this lens correction tool.
 
Threw together a slightly higher resolution map and ran the sightlines method agian on the Carmen Taylor and Michael Hezarkhani images. Camera location was again fairly close to Femr2's although it's hard to be precise since his grid only contains WTC 1 and 2 and the foreground buildings and needed some scaling and rotation to get a reasonable match. I did notice Femr2's WTC buildings are rotated about 2.5° clockwise compared to my map. The results were similar to the lo res check and again, nothing to suggest the 12° figure is too far off. Since Femr2 argued that the method is flawed when analysing 3d imagery (or at least 2d imagery of 3d space) I thought I'd test it out against a shot from much closer with a higher look up angle. Again the method results in pretty accuate matching of sightlines across the plane of the image so unless the method gives a similar degree of error regardless of the viewing angle I'm not seeing anything that would suggest an 8-9° error.

carmentaylorsightlinesNIST12_zpsb4f699b5.jpg


closeviewhighangle_zpsb96ac229.jpg
 
Last edited:
Yes, everything measured is wrong.
:) You're backing yourself into a corner rapidly.

The only logical threshold is error=0. That's unattainable therefore everything is incorrect.
LOL. One wonders why, given the fact that you have stated "I won't put a threshold on the error to consider something correct or not." why you are pursuing error propogation quantification :eye-poppi

Talk about shooting yourself in the foot :)

I did so to: Emphasize that the issue at hand is precision, not correct vs. incorrect.

If you ask once again the response won't vary the next time.
So exactly how does changing values provided in degrees to values in radians achieve that ? :confused:

Look at my response in context.
I have, and you are digging yourself into a deep hole rather rapidly.

Those of us who can interpret things in their context will understand that the values in this whole paragraph are in radians
ROFL. I'm fully aware of the units. See above, re "why" change from degrees.

(to a normal person)
:rolleyes:

you miss the context
Nope.

But that's your fault, not mine.
Nope.
 
LOL. One wonders why, given the fact that you have stated "I won't put a threshold on the error to consider something correct or not." why you are pursuing error propogation quantification :eye-poppi

Makes sense to me. pgimeno thinks that "correct" is a red herring.* That isn't a reason not to quantify error propagation. Why would it be?

*ETA: Oh, maybe I just replicated the exact source of the confusion. "Correct" is a red herring as applied to a measurement ("is that value correct?") -- not necessarily as applied to a substantive conclusion ("was that corner's acceleration > g?").

As for the comment about 0.00001 degrees, I thought and think that that was in a similar vein. Even an essentially perfect measurement wouldn't produce correct simulation results (if we could even run the sims). That doesn't mean that we shouldn't try to (1) measure as well as we can and (2) place bounds on the measurement error.

Given that your approach on one hand, and NIST's and Reactor drone's on the other, produce such disparate results, it ought to be possible to make some headway in resolving -- or at least describing! -- that technical debate.
 
Last edited:
I just assume it is. Maybe it isn't
Indeed.

Most of your laughs have been proven unjustified.
Incorrect.

I don't know. I see nothing clearly wrong there.
I know. But you will get there at some point. It's rather fundamental.

Have a serious think about what you are trying to do.

Do you mean that a small difference in FOV can have enormous effects in lens distortion?
No.

Whilst I have prompted you to understand lens distortion in more depth, I have been asking you to think carefully about your responses for a very specific and separate reason. To give you a massive hint...at the correction end of things.

I've warned you that going down this road is a can of worms, so...

Here are the two views in question
Yeah, whatever...

and here's the comparison with the corrected zoomed view, still using Gimp 2.6 built-in lens correction tool, which only has two parameters and center adjustment
What algorithm does GIMP 2.6 use to apply lens distortion correction ?
 
Oh, maybe I just replicated the exact source of the confusion. "Correct" is a red herring as applied to a measurement ("is that value correct?") -- not necessarily as applied to a substantive conclusion ("was that corner's acceleration > g?").
pgimeno's position seems to shift, depending upon his intent at the time.

If he feels like performing an error analysis for WTC7 acceleration, good luck to him.

He has a few obstacles to overcome, but hey...
 
Last edited:

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom